i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/he

Membrane reactors for green hydrogen production

from biogas and biomethane: A techno-economic
assessment

Michele Ongis a,b,*, Gioele Di Marcoberardino c, Giampaolo Manzolini a,

Fausto Gallucci b, Marco Binotti a,**
a
Politecnico di Milano, Dipartimento di Energia, via Lambruschini 4a, 20156, Milano, Italy
b
Eindhoven University of Technology, Department of Chemical Engineering and Chemistry, Sustainable Process
Engineering, De Rondom 70, Eindhoven 5612 AP, the Netherlands
c
Universita
degli Studi di Brescia, Dipartimento di Ingegneria Meccanica e Industriale, via Branze 38, 25123, Brescia,
Italy

highlights

Small-scale hydrogen production process intensification with membrane reactors.

Biogas and biomethane autothermal reforming are compared.

Detailed model of the fluidized bed membrane reactor and of the overall process.

Optimization of reactor and system performance and of levelized cost of hydrogen.

article info abstract

Article history: This work investigates the performance of a fluidized-bed membrane reactor for pure
Received 26 August 2022 hydrogen production. A techno-economic assessment of a plant with the production ca-
Received in revised form pacity of 100 kgH2/day was carried out, evaluating the optimum design of the system in
21 January 2023 terms of reactor size (diameter and number of membranes) and operating pressures.
Accepted 25 January 2023 Starting from a biomass source, hydrogen production through autothermal reforming of
Available online 1 March 2023 two different feedstock, biogas and biomethane, is compared.
Results in terms of efficiency indicates that biomethane outperforms biogas as feed-
Keywords: stock for the system, both from the reactor (97.4% vs 97.0%) and the overall system effi-
Hydrogen production ciency (63.7% vs 62.7%) point of views. Nevertheless, looking at the final LCOH, the
Biogas additional cost of biomethane leads to a higher cost of the hydrogen produced (4.62
Biomethane V/kgH2@20 bar vs 4.39 V/kgH2@20 bar), indicating that at the current price biogas is the more
Membrane reactors convenient choice.
Process intensification © 2023 The Authors. Published by Elsevier Ltd on behalf of Hydrogen Energy Publications
Modelling LLC. This is an open access article under the CC BY license (http://creativecommons.org/
licenses/by/4.0/).

* Corresponding author. Politecnico di Milano, Dipartimento di Energia, via Lambruschini 4a, 20156, Milano, Italy.
** Corresponding author.
E-mail addresses: michele.ongis@polimi.it (M. Ongis), marco.binotti@polimi.it (M. Binotti).
https://doi.org/10.1016/j.ijhydene.2023.01.310
0360-3199/© 2023 The Authors. Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5 19581

Nomenclature rj Reaction rate of reaction j. kmol/(h$kgcat)

Si Size of component i. kg or m2 or kW or L/h
A Reactor cross section. m2reactor
Si;0 Reference size of component i. kg or m2 or kW or L/
Ci Cost of plant component i. kV
h
Ci;0 Reference cost of plant component i. kV
T Reactor temperature. K
Ci;b Concentration of chemical component i in bubble
u Superficial velocity of the gas crossing the reactor.
phase kmol/m3reactor
mreactor/s
Ci;e Concentration of chemical component i in
umf Gas velocity to maintain the bed in minimum
emulsion phase kmol/m3reactor
fluidization conditions. mreactor/s
CEPCIy Chemical Engineering Plant Cost Index at year y
Waux Auxiliaries consumption. kW
dmembrane Diameter of a single membrane. mmembrane
z Axial position along the reactor. mreactor
Ea;j Activation energy of reaction j. kJ/mol
Ea;perm Activation energy term in permeance calculation. Greek Letters
kJ/mol db Bubble-phase fraction.
FH2 ;perm Total hydrogen permeated through membranes DT Temperature difference. K
per unit length. kmol/(h$mreactor) hel;ref Average electric efficiency of the power generating
Fi Molar flow rate of component i along the reactor park. %
kmol/h nj;i Stoichiometric coefficient of component i in
heq Equivalent hours of the plant. h/y reaction j. -
JH2 ;perm Hydrogen permeated per unit of membrane area rcat Apparent density of catalyst particles packed in
kmol/(h$mmembrane$ mreactor) the reactor kg/m3reactor
Kbe Exchange coefficient between bubble and
Acronyms
emulsion phase. 1/h
ACM Aspen Custom Modeler®
Keq;j Equilibrium constant of reaction j. barx
AP Aspen Plus®
k0j Pre-exponential factor of reaction j. kmol/
BG BioGas
(h$kgcat$barx)
BM BioMethane
mF Total mass of fuel (BM or BG) fed to the system. kg/
CCF Capital Charge Factor
s
CFD Computational Fluid Dynamics
mH2 Total mass of pure hydrogen produced. kg/s
C&OC Owner's and contingencies costs
n Exponential parameter in Richardson's law. -
FBMR Fluidized Bed Membrane Reactor
nCH4 ;in Moles of methane fed to the membrane reactor
GHGs GreenHouse Gases
kmol/h
HRF Hydrogen Recovery Factor
nH2 Total moles of pure hydrogen produced. kmol/h
IC Indirect costs
Nmembranes Number of membranes inserted in the
IEA International Energy Agency
membrane reactor.
LCOH Levelized Cost Of Hydrogen, V/kg
nO2 ;in Moles of oxygen fed to the membrane reactor
LHV Lower Heating Value, kJ/kg
kmol/h
MR Membrane Reactor
NR Number of chemical reactions involved
O&M Operation and Maintenance costs, V
p Pressure. bar
SEC Specific Energy Consumption, kWh/kg
Per0H2 Permeance pre-exponential factor of hydrogen
TIC Installation costs
through the membrane. kmol/
TPC Total Plant Cost, kV
(h$mmembrane$mreactor $barn)
TRL Technology Readiness Level
pi Partial pressure of component i. bar
R Universal gas constant. kJ/(mol$K)

to 24% in 2040. Despite this improvement, these numbers

Introduction
Decarbonization of the energy sector plays a fundamental role
in the reduction of greenhouse gases (GHGs) emissions. Many
results have been achieved in recent years for clean electricity
production by low-carbon sources and the total share of
electricity in global final energy consumption is growing. Net
Zero by 2050 scenario of the International Energy Agency (IEA)
claims a share of almost 50% electricity on total energy con-
sumption in 2050 [1]. In the stated policies scenario, reported
again by IEA [2], the share of electricity goes from 19% in 2018
indicate that the transition of the electric sector only cannot
be enough on its own in the overall GHGs emission reduction,
particularly since global energy demand is constantly
growing. In the stated policies scenario, world energy demand
for 2040 is expected to increase by 50% compared to the value
in 2019 [2]. Since electrification only cannot bear the entire
burden of decarbonization, multiple fuels and technologies
will be needed in this process. Among them, a critical role in
the hard-to-abate sectors, such as heating and mobility, is
played by low-carbon gaseous carriers, led by biogas, bio-
methane and low-carbon hydrogen. The potential diffusion of
19582 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
hydrogen-based systems, such as fuel cell and industrial
boilers, will result in a higher green H2 demand.
Today, dedicated hydrogen production is about 72 million
tonnes (Mt) every year. It is almost entirely produced by fossil
fuels, mainly from steam methane reforming process, and
implicates significant GHGs emissions, around 830 Mt of CO2
every year [3]. To produce low-carbon hydrogen, the primary
route is the usage of fossil fuels with carbon capture, utiliza-
tion and storage, accounting for 0.9% of hydrogen production
in 2020, followed by water electrolysis [4]. Another method to
produce hydrogen from renewables sources is the usage of
biomass, mainly through pyrolysis and gasification [5].
Emerging technologies for hydrogen production from biomass
involves membrane technology, starting from biogas (BG) or
biomethane (BM) obtained through biomass anaerobic diges-
tion. Promising results for a single-step production, suitable
particularly for distributed applications, have been recently
obtained both with active proton-conductive ceramic mem-
branes in proton ceramic reactors [6] and with Palladium-
alloys based (Pd-based) membranes in membrane reactors
[7]. In this work, the pathway investigated is green hydrogen
production through steam methane reforming of BG or BM in
membrane reactors containing Pd-based membranes.
Biogas is basically a mixture of methane and carbon diox-
ide, with traces of other gases (O2, N2, H2, H2O, H2S, NH3). It is
mainly produced by anaerobic digestion in biodigesters, but
can also be originated from landfill gas recovery systems or
wastewater treatment plants. Biomethane is the renewable
form of the natural gas, mainly composed by methane. About
90% of biomethane is produced from raw biogas, through a
post-treatment process called biogas upgrading, while the
Fig. 1 e Top: concept of membrane reactor used for hydrogen production process intensification; Bottom: paths investigated
in this work.
remaining part is directly produced from solid biomass gasi-
fication followed by methanation [8]. In upgrading processes,
carbon dioxide and other contaminants are removed from
biogas, obtaining nearly pure methane (>98%) [9]. Mostly used
processes for the upgrading are water scrubbing, chemical
scrubbing, pressure swing adsorption and membrane tech-
nology. The choice of a particular process depends on many
factors, such as the methane loss (from 0.04% to 4%), biogas
pre-cleaning requirement, difficulties in control of the pro-
cess, heat requirement, purity of produced biomethane and
investment and operational costs. In Europe, most of the
biogas is used for power generation, heat in buildings and
cogeneration units, while only about 10% is upgraded to bio-
methane [8]. Nevertheless, it is progressively observed a shift
towards upgrading, mainly due to advancement in upgrading
technology, subsidies and new opportunities for biomethane
in natural-gas powered vehicles and for its injection in the
natural-gas grid [10]. Biogas production expected potential for
2040 is 50% larger than today [8]. Moreover, its production
average cost is expected to fall over time. Its increasing
availability, lower production cost and increasing in upgraded
share to biomethane share make these two feedstocks inter-
esting for green hydrogen production.
Process intensification via membrane reactor (MR) tech-
nology has been proposed as a promising option to reduce
costs and increase efficiency for small-scale hydrogen pro-
duction. By integrating selective membranes in the reactive
zone of the reactor, pure hydrogen is produced in a single
equipment, thus avoiding the expensive downstream pro-
cesses. This concept is illustrated in Fig. 1. Moreover, products
removal in the reactive zone allows to push forward chemical
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
equilibrium limitations of reactions involved, leading to higher
conversion compared to a conventional reformer reactor. MRs
have shown interesting potentialities in different EU funded
projects [11]. Particularly relevant for hydrogen production
from biogas was BIONICO project [12], where a prototype of a
Fluidized-Bed Membrane Reactor (FBMR) with the capacity of
100 kg/day was designed and built. Currently, the EU project
MACBETH [13] aims to bring this technology to TRL7 by 2024 for
different processes, including hydrogen production.
In this context, modelling activity is a fundamental support
to the industrial development of MRs technology. This is
particularly true in fluidized-bed configuration, where a com-
plex fluid dynamics interacts with the reactive and the
permeation process. A multi-scale modelling is in general
necessary to have a precise description of the process. Perme-
ation models [14] and kinetic models [15] have to be developed
or selected from literature to be included in the FBMR model. At
reactor level, detailed Computational Fluid Dynamics (CFD) and
optical techniques can be used for a detailed description of
some fundamentals about the fluidized bed behaviour [16e18].
The Reactor model should have a level of detail coherent with
the aim of the analysis, being as much as possible consistent
with the results of the fundamentals models. FBMR models in
literature have been gradually developed to refine the
phenomenological description [19e21]. At system level, reactor
model is included in a broader process simulation to perform a
techno-economic analysis of the overall system [22e25].
In this work, the use of a FBMR for pure hydrogen produc-
tion through an autothermal reforming process is investigated.
The size chosen is 100 kg/day, coherently with the prototypes
developed within BIONICO and MACBETH, which represents a
typical distributed hydrogen production plant for refuelling
stations, glass industries and other small industries. In a pre-
vious work [23], the potentialities of the system had already
been investigated by the identification of the maximum effi-
ciency operating points together with a preliminary economic
analysis using raw biogas as feedstock. Nevertheless, it was
limited to the reactor diameters and membrane areas that led
to the maximum system efficiency. Purpose of the present
work is to perform a techno-economic assessment to compare
biogas and biomethane as feedstocks for hydrogen production
in a FBMR, based on the analysis performed in Ref. [23], with
the following improvements: (i) the optimization is performed
with an enhanced version of the model, showing clearly the
efficiency trends obtained by varying the manipulated vari-
ables and (ii) the economic analysis is extended to all the
working points investigated, since the economic optimum can
in general be different from the technical optimum.
The use of biomethane, avoiding to feed carbon dioxide to
the membrane reactor as in the biogas case, leads to higher
partial pressure of the hydrogen in the reactor, thus
increasing the driving force required for permeation. This in
principle allows a reduction of the number of membranes
required and of the reactor size, thus reducing the hydrogen
production cost. On the other hand, biomethane production
requires an upgrading section that can raise the total cost.
Results are compared looking at reactor and system effi-
ciencies and at the hydrogen cost.
General methodology is presented in section 2, together
with the models assumptions and plant layout. The simulation
19583
results are presented in section 3. Section 4 describes the
economic analysis, followed by conclusions in section 5.
Methodology
This work is based on a detailed mathematical model of the
membrane reactor. The model has been realized in the soft-
ware Aspen Custom Modeler® (ACM), which includes a data-
base for the properties of the chemical components involved
in the process and allows a direct integration with Aspen
Plus® (AP), the software used to perform the simulation of the
overall process.
In section 2.1, the general assumptions of the work will be
listed, as well as the key performance indicators used to
evaluate the performance from the point of views of both
reactor and overall plant. The general methodology of the
techno-economic assessment is mainly taken by Ref. [23], at
which most of the assumptions refer to. In section 2.2 and 2.3,
methodology of respectively the reactor model and the overall
system model are presented.
General assumptions
Both ACM and AP allow to call various set of properties of the
chemical components involved in the process. The equation
of state selected for the computation of all the thermody-
namic properties is the Peng-Robinson, widely used in pro-
cesses involving hydrocarbons.
The feedstock of the process is biogas or biomethane
depending on the case studied (BG case or BM case). BG
composition is in general variable and depends on the
biomass source: the one used in this work is taken from
Ref. [23] and is reported in Table 1. BM composition, reported
in Table 1, is taken from Ref. [26]. Lower Heating Values (LHVs)
are provided by the software AP.
In the investigated cases, hydrogen is produced from
biogas or biomethane reforming in an autothermal process.
Considering all the chemical reactions involved, the chemical
components present in the system are CH4, H2O, CO, CO2, H2,
O2, N2. H2S is not considered as membranes and reforming
catalyst would not withstand the presence of H2S [27] and thus
this component needs to be separated before the reactor (as it
happens in conventional reforming).
Reactor type and reactor model
The FBMR is a cylindrical vessel with the membranes verti-
cally inserted from the top part of the reactor. In this work, the
Table 1 e Biogas and biomethane mole fractions (%) used
in this work.
Species Biogas Biomethane
CH4 58.1 96.0
CO2 33.9 2.0
N2 3.8 1.5
O2 1.1 0.5
H2O 3.1 e
LHV (MJ/kg) 17.8 45.7
19584 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
membranes are composed by a selective PdeAg layer plated
on a ceramic support, as described in Ref. [7], with a cylindrical
structure and an inner cylindrical space for hydrogen collec-
tion where vacuum is obtained through a vacuum pump. Re-
actants (feed) are fed from the bottom of the reactor and flow
towards the top. Hydrogen collected in the inner part of the
membrane is called permeate, while the chemical components
which flows towards the top without crossing the membranes
compose the retentate. The gas flowing in the reactor should
have enough superficial velocity to drag the catalyst particles,
bringing the reactor in bubbling fluidization regime. From the
modelling point of view, the reactor can be divided in three
different regions: the bottom part, where only catalyst is
present (no membranes), is a pure reactive region, as in a
conventional fluidized bed reactor (this allows producing
hydrogen and effectively using the membranes afterward);
the central part, once membranes start, in which both
chemical reactions and permeation occur; at the top of the
reactor, above the membranes, there is the freeboard region
where in general reactor section increases to slow down the
retentate gas, allowing the solid catalyst particles to fall down
again in the region where membranes operate. Reactor
structure is represented in Fig. 2.
The model describes the first two sections: the reaction
zone and the permeation zone, then assuming that the pres-
ence of catalyst particles in the freeboard region does not lead
to an additional conversion. It is a 1D model, discretized along
the vertical length of the reactor, here referred as direction z.
The fluidized-bed is modelled according to the two phase
theory: bubble phase and emulsion-wake phase [28]. Bubbles are
spherical and their diameters - increasing with height - as well
as the total amount of bubbles into the reactor, are evaluated
from correlations obtained for a fluidized-bed reactor without
membranes [29], since no accurate correlations for fluidized-
bed with immersed membranes are available yet. The ACM
code of the model, as well as its mathematical description, is
based on a previous work [21], where a detailed description of
the model as well as all the correlations used are reported,
Fig. 2 e Schematic of the FBMR structure. In red, the
sections considered in the model. (For interpretation of the
references to color in this figure legend, the reader is
referred to the Web version of this article.)
including calculations for minimum fluidization velocity, void
fraction, bubble diameter, bubble and emulsion phase frac-
tions, wake fraction and exchange coefficients between the
phases. In this model it is assumed that no solid particles are
present in the bubble phase, therefore no reactions occur in-
side them. The bottom part of the bubble is called wake and
contains solid particles. Wake fraction is calculated starting
from bubble fraction by a correlation taken from Ref. [17].
Emulsion phase is assumed to be in minimum fluidization
conditions and is the phase where chemical reactions happen.
For simplicity, also the wake is assumed to be in minimum
fluidization condition. Accordingly, emulsion and wake pha-
ses can be considered as a unique emulsion-wake phase. The
excess gas flows as bubbles inside the bed. Moreover, it is
assumed that hydrogen permeation through the membranes
occurs from both bubble and emulsion phases according to
their fraction.
The model performs material balances in each element dz
and an overall energy balance to verify the autothermal
behaviour of the reactor (assuming uniform temperature in-
side the bed, as experimentally verified for several fluidized-
bed reaction systems). The material balance is performed for
each element dz, for each chemical component and for each
phase, and is structured as reported in Equation (1).
vFi ðzÞ X
NR
 
¼ A,rcat , rj ðzÞ,nj;i ± A,db ðzÞ,Kbe ðzÞ, Ci;b ðzÞ  Ci;e ðzÞ
vz j
 FH2 ;perm ðzÞ
Equation 1
Each balance considers the variation in the molar flow of
each component Fi from the inlet to the outlet of each
element, depending on different terms: the first term con-
siders the component production/consumption in chemical
reactions, calculated for each reaction j of the total number of
reactions taken into account, NR. In this work, the reactions
included in the model are methane steam reforming (R.1), water
gas shift (R.2) and methane oxidation (R.3).
CH4 þ H2 O 4CO þ 3H2 R.1
CO þ H2 O 4CO2 þ H2 R.2
CH4 þ 2O2 / CO2 þ 2H2 O R.3
The chemical reaction term depends on the mass of cata-
lyst at each axial position z, given from the product of catalyst
apparent density rcat (that is, particle density multiplied by
reactor void fraction) and the reactive area A, the reaction rate
of each reaction (rj;i ) and the stoichiometric coefficient of the
component i in the reaction j (nj;i ), considered positive for
products, negative for reactants and zero for components not
involved in that reaction. For reactions R.1 and R.2, reaction
rates are taken from Ref. [30], while their parameters are taken
from Ref. [15] and reported here in Equation (2) and Equation
(3) respectively. Reaction rates are calculated from partial
pressure of reactants and products pi and from an Arrhenius-
type coefficient including a pre-exponential factor (k0j ) and an
exponent which considers the activation energy (Eact;j ) over the
product between universal gas constant (R) and reactor
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5 19585

temperature (T). For R.3, since methane oxidation is in general

Table 2 e Reactor, membranes and catalyst parameters.
much faster than the other two reactions and it resulted from
kinetic modelling that all oxygen in consumed within the first Parameter Value Units
5 cm (where membrane region start), it is assumed that all Reactor length 50 cm
oxygen fed to the reactor instantly oxidises methane at the Membrane length 45 cm
reactor inlet, to reduce computational efforts and improve Membrane diameter 1.4 cm
Distance between membrane 3.4 cm
convergence.
centers
   p3 ðzÞ,pCO ðzÞ
 Density solid particle 2000 kg/m3
Ea;SMR
k0SMR ,exp R,T , pCH4 ðzÞ,pH2 O ðzÞ  H2Keq;SMR Diameter solid particle 180 mm
rR:1 ðzÞ ¼ k0SMR 9:43,105 kmol=ðh,kg,bar0:404 Þ
H2 O ðzÞ
p1:596
Ea;SMR 106.9 kJ/mol
Equation 2 k0WGS 8:82,102 kmol=ðh,kg,barÞ
Ea;WGS 54.5 kJ/mol
   p ðzÞ,pCO2 ðzÞ
 Per0H2 9.94 kmol=ðh,m,bar0:5 Þ
Ea;WGS
k0WGS ,exp R,T , pCO ðzÞ,pH2 O ðzÞ  H2 Keq;WGS
rR:2 ðzÞ ¼ Ea;perm 9.26 kJ/mol
pH2 O ðzÞ n 0.5 e
Equation 3

The kinetic term is zero if the balance is applied to the

bubble phase or if an inert component, such as nitrogen, is
considered. The total amount of solids is computed directly by
the model, once specified particle density and diameter, by
means of a correlation which relates the void fraction in
minimum fluidization condition and the fluid dynamics, re-
ported in Ref. [21]. A fraction of the total amount of solids is
constituted of catalyst particles (Nickel or Rhodium formula-
tion over an alumina support), while the remaining part is
constituted of inert alumina support particles. Both types of
particles are assumed to have the same density and diameter,
as experimentally verified. Values used in this work are for a
Ni-based catalyst, but, considering that catalyst amount
assumed is enough to reach maximum conversion, similar
results can be obtained with a lower amount of Rh-based
catalyst. The ratio between catalytic particle and filler parti-
cle is set to 50% for Ni-based catalyst, while 1/10 is sufficient
for Rh-based catalyst, considering its high activity.
The second term in Equation (1) is the flux between the
bubble and the emulsion phase, driven by concentration dif-
ference in bubble (Ci;b ) and emulsion (Ci;e ). In this relation, the
plus holds when the balance is performed on the emulsion
phase, while the sign is minus if applied to the bubble phase. db
is the bubble fraction in the bed and Kbe the exchange coeffi-
cient: both terms depend on fluidization conditions and are
estimated using correlations reported in Ref. [21]. The last term
of Equation (1) is hydrogen permeated for unit length (FH2 ;perm )
at each axial position z and should be included only when the
mass balance is applied to component hydrogen. The amount
of hydrogen permeated for unit membrane area in each
element (JH2 ;perm ) is calculated using the Richardson's equation,
reported in Equation (4), where the driving force is the differ-
ence in hydrogen partial pressure between retentate side of the
membrane (pH2 ;retentate ) and permeate side of the membrane
(pH2 ;permeate ). Flux depends on an exponent n and an Arrhenius-
type permeance, characterized by a pre-exponential per-
meance coefficient (Per0H2 ) and an activation energy (Ea;perm ),
that can be derived from permeation experiments.
 
Ea;perm  n
JH2 ;perm ðzÞ ¼ Per0H2 ,exp , pH2 ;retentate ðzÞ  pnH2 ;permeate ðzÞ
RT
Equation 4
This term is related to the last term of Equation (1) from the
following relation (Equation (5)), which express the hydrogen
permeated for unit of membranes length, considering the
number (Nmembranes ) and diameter of the membranes (dmembrane ).
FH2 ;perm ðzÞ ¼ Nmembranes ,p,dmembrane ,JH2 ;perm ðzÞ Equation 5
Geometric parameters of the reactor and of the mem-
branes, as well as kinetic parameters for reaction rates
equations and membrane parameters for permeation behav-
iour, are reported in Table 2.
Compared to the previous works, here the algorithm and
the fluidization model have been improved. Mainly, the ordi-
nary differential equations convergence have been re-written
using ACM tools to optimize the solutions and the accuracy of
the solution was refined increasing the reactor discretization
and the order of convergence of the differential equations
implemented in the model, assuring a good compromise be-
tween convergence time and grid independent results. The
decisions of the number of grid points came from preliminary
analysis, performed with different order of convergence and
convergence method. Firstly, the amount of hydrogen
permeated through the membrane is calculated at different
grid points, using the highest level of accuracy, provided in
ACM by a UPwind-based Backward-Difference method of the
4th order (UPBD4). Results are reported on the left of Fig. 3, and
shows that the results on the amount of hydrogen permeated
are grid independent above 500 points. For a lower number of
points, production is slightly underestimated (about 1% less
working with 100 points).
To reduce the computational effort, an additional analysis
investigated the results with different convergence methods.
The comparison was between three finite-difference methods
with different order of convergence (BDF1, BDF2, UPBD4, all
based on backward difference) and a finite-elements method
(OCFE2). Last number in the name indicates the order of
convergence. Results, reported on right side of Fig. 3, showed
that above 250 points all methods guaranteed a reasonable
accuracy (below 0.5%), with the error halved going from 1st to
 2nd order method and substantially unchanged for higher
orders. Thus, it has been decided to use 250 grid points and a
2nd order method (BDF2).
19586 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5

Fig. 3 e Preliminary analysis on equations convergence to select the number of grid points and the method and order of
convergence. Results on the left are obtained with UPBD4 method, which guarantees maximum accuracy. (For
interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Regarding the bubble-emulsion phase behaviour, the

description of the bubble wake was added (wake phase). As in
the previous version, the model neglects concentration po-
larization losses, because no correlations are available for
fluidized bed with vertical membranes and, according to
experimental test on lab test prototype, it can be expected
their influence is lower at large scale compared to smaller
reactors. Possible mass transfer limitations in the membrane
support are also neglected.

System layout and assumptions

The system layout for both the investigated feedstocks is re-

ported in Fig. 4. Compressed BG/BM is mixed with compressed
air and then preheated up to maximum 300  C to avoid
methane cracking. Just before reactor inlet, biogas-air mixture
is mixed with steam and enters the reactor. The steam is
produced from water at ambient conditions, pumped until
reactor pressure and heated in three heat exchangers (HX-1,
HX-2, HX-3), reaching a temperature above 500  C, to end up
with a feed temperature of about 400  C at the reactor inlet
and thus avoiding strong thermal gradients within the
reactor. Inside the reactor, the oxidation reaction will rise the
temperature up to 550  C and, in addition, will provide the
heat required by the reforming reactions. In the fluidized-bed
reactor, no temperature gradient is considered and an
isothermal profile along z is assumed. Permeate flow, con-
sisting of pure hydrogen, and retentate flow are then both at
550  C. Permeate is cooled down in two heat exchangers: the
first one provides heat for to steam superheating (HX-3), and
the second one to the water preheating (HX-1). Once cooled
down, the permeated hydrogen is led to atmospheric pressure
by two-stages refrigerated vacuum pump and then com-
pressed with an intercooled compressor up to 20 bar. This
choice of the delivery pressure has been done to be consistent
with previous works and has no effect in the BG/BM cases
comparison since pure hydrogen production is fixed, and then
the amount of energy required for H2 compression is the same
Fig. 4 e System layout.
in both cases. The retentate flow is cooled down in HX-2 to
provide the heat for steam production, then it is throttled to
atmospheric pressure and sent to a catalytic burner to com-
plete the oxidation of CO, H2 and CH4. These reactions provide
additional heat to the flue gases, used for the BG/BM pre-
heating (up to 300  C) and for the air pre-heating.
Manipulated variables are the feed pressure, the reactor
diameter and the number of membranes inserted in the
reactor. Feed pressure investigated are 10, 12 and 14 bar,
consistently with the results obtained in a previous work for
the same system [23]; reactor diameters investigated go from
36 cm to 45 cm, since it is a range that ensure correct fluid-
ization and correct fitting of all the membranes. The number
of membranes reported is from 70 to 134, since for lower
values the drop in efficiency is in all cases very relevant and
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5 19587

Table 3 e Example of main streams of BG case with 120 membranes.

Flow Composition (% molar basis)
Stream T ( C) p (bar)
Molar (mol/s) Mass (g/s) CH4 H2 CO CO2 H2O O2 N2
1 0.333 8.717 50 1 58.1 0 0 33.9 3.1 1.1 3.8
2 0.292 5.266 520.5 10 0 0 0 0 100 0 0
3 1.032 25.725 363.3 10 18.7 0 0 10.9 29.3 8.6 32.4
4 0.753 24.568 198.8 10 0.1 1.2 1.1 39.4 13.8 0 44.4
5 1.001 32.156 333.5 1 0 0 0 30.4 11.4 4.5 53.8
6 0.574 1.157 141.3 20 0 100 0 0 0 0 0

Table 4 e Example of main streams of BM case with 70 membranes.

Flow Composition (% molar basis)
Stream T ( C) p (bar)
Molar (mol/s) Mass (g/s) CH4 H2 CO CO2 H2O O2 N2
1 0.203 3.424 50 1 96.0 0 0 2.0 0 0.5 1.5
2 0.319 5.751 324.7 14 0 0 0 0 100 0 0
3 0.943 21.305 339.3 14 20.7 0 0 0.4 33.9 9.5 35.6
4 0.664 20.148 163.6 14 0.4 1.7 0.9 28.7 17.9 0 50.5
5 0.902 27.249 378.8 1 0 0 0 22.1 15.0 4.2 58.7
6 0.574 1.157 141.3 20 0 100 0 0 0 0 0

for higher values the costs rise without additional benefits. An membranes for each case is tied to reactor diameter, since the
example of typical flow rates and compositions of the main minimum distance between membranes centers is fixed to
streams are reported in Table 3 for BG and in Table 4 for BM. 3.4 cm, and thus for smaller diameters only a limited number
Cases selected are two limit cases: for BG, main streams are of membranes fit into the reactor. The center-center distance
reported at 10 bar and with 120 membranes. For BM, at 14 bar selected derives from the consideration that at least 2 cm of
and with 70 membranes. space should be left between the membranes to avoid a strong
The assumptions for the system are mainly taken from increase of the concentration polarization losses [31], and
Ref. [23] and are reported in Table 5. The maximum number of adding the membrane outer diameter, which is 1.4 cm. To
ensure a correct fluidization of the bed, it should be verified
that the gas velocity stays in a certain range: in this work, the
range allowed of the ratio between superficial velocity u and
Table 5 e System assumptions for the simulations. minimum fluidization velocity umf is between 1.5 and 5. The
Fixed parameter Value Units lower boundary is to be conservative in avoiding defluidiza-
Reactor uniform temperature 550 
C tion. The fastest boundary is chosen to be conservative on
Vacuum side pressure 0.1 bar avoiding the removal of catalyst particles from the reaction
Steam-carbon-ratio at x ¼ 5 cm 3 e zone due to an increase in the gas drag force.
Fluidization ðu =umf Þ range 1.5 / 5 e Regarding the reactor feed inlet, BG/BM flow rate is set in
Hydrogen production 100 kg/day
order to produce the target amount of pure hydrogen, while
Hydrogen delivery pressure 20 bar
 the steam flow rate is set such that the steam-carbon-ratio,
Ambient temperature 25 C
Controller consumption 10 % defined as the ratio between steam and methane molar flows
(% of total auxiliary consumption) in the feed, is fixed at the beginning of the membrane region,
Average electric efficiency of the power 45 % to avoid carbon deposition over the catalyst and over the
generating park membrane surface. Experimental results [15] show that a
Water pump hydraulic/mechanical 0.7/0.9 e value of 3 should be enough conservative. Air flow rate to the
efficiency
reactor is tuned to obtain autothermal conditions within the
Vacuum pump isentropic/mechanical 0.7/0.85 e
reactor. On the other hand, the air flow rate to the catalytic
efficiency
Compressors isentropic/mechanical 0.7/0.85 e burner is instead set in excess to have an oxygen molar frac-
efficiency tion in the dry flue gases equal to 5%.

Minimum DT in heat exchangers 30 C
Heat transfer coefficient gas/gas 60 W/(m2 , K) Key performance indicators
Heat transfer coefficient gas/ 70 W/(m2 , K)
liquid-biphasic
To evaluate the performance and to compare the different
LHVH2 120 MJ/kg
feedstocks, mainly two indicators are used. From the reactor
Investigated parameter Value Units point of view, the efficiency parameter investigated is the
Retentate pressure 10 -12 - 14 bar Hydrogen Recovery Factor (HRF), defined as the ratio of moles
Reactor diameter 36 / 45 cm of hydrogen separated (nH2 ) through the membrane (so the
Number of membranes 70 / 134 e pure hydrogen produced) with respect to the total ideal
19588 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
hydrogen production obtained under the hypotheses that: i)
all methane fed to the reactor and available for reforming is
converted to H2 and ii) all hydrogen produced in the
reforming reactions is separated through the membranes.
HRF is defined in Equation (6). The expression for the de-
nominator becomes clear considering that, through re-
actions R.1 and R.2, for each mole of methane that
undergoes to reforming (nCH4 ;in ), 4 mol of hydrogen can be
obtained; in the meantime, 1 mol of methane is burned
every 2 mol of oxygen fed to the reactor and then the moles
of methane that do not go to reforming are half the moles of
oxygen fed (nO2 ;in ).
nH2
HRF ¼  n
 Equation 6
4, nCH4 ;in  O22 ;in
At system level, the HRF alone is no longer the best indicator
to identify the best design point and it could instead be useful to
refer to an energy efficiency of the system, defined as the energy
output associated to the hydrogen produced over the total power
inlet to the process. The energy input to this process is due to the
biogas/biomethane fed and to the primary energy necessary to
produce the electric energy for auxiliaries’ consumptions. The
system efficiency (hsystem ) and the Specific Energy Consumption
(SEC) are defined in Equation (7) and Equation (8) respectively.
LHV for hydrogen, biogas and biomethane are respectively
120 MJ/kg, 17.8 MJ/kg and 45.7 MJ/kg. The terms mH2 and mF refer
to the mass flow of hydrogen permeated and mass flow of
feedstock sent to the reactor respectively.
mH2 ,LHVH2
hsystem ¼
mF ,LHVF þ W aux
hel;ref
LHVH2 mF ,LHVF þ hel;ref
Waux
SEC ¼ ¼ Equation 8
hsystem mH2
Fig. 5 e Effect of feed pressure and reactor diameter on hydrogen recovery factor (left) and system efficiency (right) for
different number of membranes inserted in the membrane reactor.
System performance
The effect of manipulated variables (retentate side pressure,
reactor diameter and number of membranes) on reactor and
system performance was investigated in a series of analyses
both for the BG and BM cases. The trends are firstly reported
for BG case. Comparison between BG case and BM case is re-
ported in section 3.2.
System efficiency and HRF trends
Results of the analysis of the reactor behaviour are reported in
terms of HRF at the left of Fig. 5. Clearly, by increasing the
number of membranes, the HRF increases as well, since there
is more membrane area available for hydrogen to permeate.
The positive effect of an additional membrane is always lower
as the number of membranes increases, until HRF reaches an
asymptote. The increase of the reactor diameter leads also to
an increase of the reactor efficiency, and this is mainly due to
the reduction in the gas velocity due to the increased cross
section area. Reduction in gas velocity leads to a higher resi-
dence time and therefore to higher conversions. This effect is
of course limited to the fact that the fluidization regime should
be guaranteed. In the investigated range, also the effect of
pressure increase is beneficial for the reactor. In general,
reforming reactions are penalised by high pressure since they
occur with an increase of total number of moles of the system.
Nevertheless, higher pressures entail lower velocities e due to
reduction of gas density and then volumetric flow - and then
Equation 7 an increase in the residence time. Most important, in mem-
brane reactors, an increase in pressure has the beneficial effect
of increasing the hydrogen partial pressure at the retentate
side, and then the driving force for hydrogen permeation. In
this case, driving force gain effect overcomes the drop in
equilibrium conversion. Ideally, with an infinite number of
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
membranes fitted into the reactor, all the curves reaches the
same asymptotic value of HRF, obtained when hydrogen has
the same partial pressure at both sides of the membranes.
The reactor results anyhow do not take into account heat
integration and auxiliaries’ consumption of the overall plant.
Thus, they cannot by themselves identify the best working
point of the system. For this reason, also the system efficiency is
evaluated. Results are reported on the right of Fig. 5. From the
system point of view, reactor diameter influence is negligible.
The trend of pressure is reversed: in general, lower operating
pressures lead to higher system efficiencies compared to higher
pressures. This trend cannot be considered as a general rule and
it is influenced by the reactor design: in this case, below 110
membranes the efficiency slope at 10 bar become steep, while at
12 bar it happens with 85 membranes. For this reason, at about
90 membranes the curves cross each other and then below this
value the higher-pressure case has an higher efficiency. The
same discussion could be done for other pressure values. More
interesting is the other side of the chart: by increasing the
number of membranes, the curves reach an asymptotic value of
efficiency, due to the steadiness of HRF. The difference, in this
case, is that such asymptotic values depend on the operating
pressure and are higher for lower pressure: for 14 bar,
maximum value is 62.6%; for 12 bar, 63.2% and for 10 bar 63.6%.
This behaviour can be explained considering that by increasing
the number of membranes, the HRF becomes the same for all
pressures. Since hydrogen production is fixed, the equivalence
of HRF means the same amount of biogas fed to the reactor.
Then, the only different term in Equation (7) among the cases is
the power required by the auxiliaries at the denominator. The
difference is mainly due to the increased power required by air
and biogas compressors at higher pressures.
Simulations results of right-side of Fig. 5 appear wobblier
compared to left-side figure since they refer to the overall
system. In simulations at system level, different design specs
are defined in AP to match the desired constraints (set values of
hydrogen production, reactor thermal duty and steam-carbon-
ratio), as well as the minimum temperature differences in the
heat exchangers. Design specs have a certain tolerance, which
leads to a small floating of the results, which does not appear in
the curves at reactor level (left side). However, the trends ob-
tained at system level are in all cases clear.
Since the trends in terms of HRF and system efficiency lead
to different conclusions, it can be interesting to compare the
two curves of one case study, so fixing reactor diameter and
operative pressure. As already stated, for a large number of
membranes both curves reach a plateau. It is however clear
from Fig. 6 that system efficiency reaches the plateau for a
lower number of membranes. This effect holds for a relevant
number of membranes: in this example, HRF has not already
reached the plateau with 127 membranes while system effi-
ciency is about constant already at 95 membranes. The reason
for that can be found in the fact that when HRF decreases,
retentate flow is richer of unconverted methane and
hydrogen, and thus it is characterized by a higher LHV. Then,
when the retentate is oxidized in the catalytic burner, there is
an additional heat recovery which, in end, increases the
temperature of the feed that enters the reactor. In case of
Fig. 6, for example, with 127 membranes the retentate com-
bustion can provide heat to bring the reactor feed at 335  C,
19589
Fig. 6 e Typical trends of HRF (blue) and system efficiency
(red) for different number of membranes. Results obtained
at 12 bar and 44 cm of reactor diameter. (For interpretation
of the references to color in this figure legend, the reader is
referred to the Web version of this article.)
while with 85 membranes it can reach 414  C. This allows to
save a certain amount of air fed to the reactor and then to
decrease the air compressor power duty (4.9 kW with 127
membranes vs 4.6 kW with 85 membranes).
Comparison between biogas and biomethane
Same charts presented above can be obtained using bio-
methane as feedstock of the system. In this case, the absence
of inert carbon dioxide in the reactor feed increases the
hydrogen molar fraction then the driving force for perme-
ation, depending on its partial pressure. In general, the effect
of manipulated variables are the same already reported for
biogas. It is then only relevant a comparison with biogas re-
sults. The results, shown in Fig. 7, clearly show that bio-
methane outperforms biogas both from the point of view of
the reactor and the overall system. In terms of HRF, at fixed
reactor diameter and number of membranes, the two positive
effects of using biomethane instead of biogas, driven by the
absence of nitrogen, are the increase in hydrogen partial
pressure and the reduction of fluidization velocity. The latter
effect is due a reduction of the molar (and then volumetric)
flow required, since the analysis is performed at fixed
hydrogen production. The advantages in using biomethane
are reduced by increasing the number of membranes: with 121
membranes and 43 cm reactor, for example, HRF is 98.7% for
BM and 98.1% for BG. With same diameter and 85 membranes,
HRF for BM is 98.1% while for BM 96.0%. A similar trend was
observed for the other pressures investigated, with the same
effect already reported for biogas case.
As discussed in the previous section, an increase in HRF is
not obviously related to an increase in system efficiency. In this
case however, biomethane outperforms biogas also from the
point of view of the overall system. Asymptotic value at 14 bar,
for example, is 63.8% for BM and 62.8% for BG, as shown at the
right side of Fig. 7. The reason for its higher value is mainly the
electric energy saving in the power required for BM compres-
sion. In the asymptotic region at 14 bar, BG compressor requires
4.4 kW while BM compressor only 2.7 kW.
19590 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
Fig. 7 e Comparison between biogas and biomethane in terms of HRF (left) and system efficiency (right) for different reactor
diameters and number of membranes at 14 bar.
From a pure technical point of view, a design criterion for
the reactor can be to work at the beginning of the plateau of
system efficiency, where the increase in membrane numbers
can potentially reduce the reactor reliability, increase the in-
vestment and maintenance costs without any additional
benefit on the overall system. So the biomethane case would
lead to 85 membranes with an efficiency of about 63.7%, while
biogas case required 95 membranes with an efficiency of
62.6%. Nevertheless, a better criterion will be given from the
economic analysis in section 4, to evaluate if it is worth to
trade some efficiency to work with lower membranes, then
having a feedstock extra cost but a reduction in membranes
cost.
Economic analysis
An economic analysis is carried out to estimate the hydrogen
production cost using both the feedstocks. The approach and
the main assumptions are mainly taken from Ref. [22], and it
consists in the evaluation of the Levelized Cost Of Hydrogen
(LCOH). Assumptions for standard components are consistent
with [22], while assumptions for the membrane reactor are
consistent with [23].
Methodology and assumptions
For the economic analysis, the parameter identified to opti-
mize the process is the LCOH, evaluated in V/kg, defined as in
Equation (9). The evaluation of the cost of hydrogen production
is composed by the total plant cost (TPC) and the Operations
and Maintenance (O&M), fixed and variables, divided by the
mass flow rate of hydrogen produced. The TPC is converted in
an annual operating cost using the Capital Charge Factor (CCF)
methodology. 7500 working hours (heq ) are considered.
TPC,CCF þ O&Mfix þ O&Mvar ,heq
LCOH ¼ Equation 9
mH2
The TPC is calculated starting from basic components cost
(Ci ) and then adding installation costs (TIC), indirect costs (IC)
and owner's and contingencies costs (C&OC), according to
Equation (10). Basic component costs were available for some
reference sizes and were scaled up using CEPCI index method,
as in Equation (11), based on their actual size (Si ), starting from
a reference cost (Ci;0 ) at the reference size (Si;0 ).
!
X
TPC ¼ Ci ,ð1 þ %TIC Þ,ð1 þ %IC Þ,ð1 þ %C&OC Þ Equation 10
i
 f !
Si CEPCI2020
Ci ¼ Ci;0 , , Equation 11
Si;0 CEPCIy
y
The economic purpose is clearly to identify the operating
conditions and reactor geometries which minimize the LCOH.
In this work, heat exchangers cost has been refined, such
as the calculation of heat exchangers area, based on the UA
value provided by AP and then assuming typical values for
heat transfer coefficient U to determine heat exchanger area,
reported in Table 5. BM cost have been taken from Ref. [8] as
the world average value, and is 19 $/MBtu, corresponding to
0.532 V/Nm3. BG cost is then estimated considering an average
upgrading cost of 3 $/MBtu [32] and subtracting it to BM cost. It
results 16 $/MBtu, corresponding to 0.2712 V/Nm3, and it is
consistent with [8]. Due to the simplicity of the system, labor
cost estimated is 30 kV/y.
CEPCI index have been updated to the 2020 value, which is
596.2 [33]. Main economic assumptions, together with the
parameters for the calculation of TPC of Equation (10), are
reported in Table 6.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5 19591

reducing reactor diameter also means that less membranes

Table 6 e Main economic assumptions on O&M, TPC and
can be fitted into the reactor.
feedstock costs.
The LCOH curve shows a parabolic trend if plotted as a
Parameter cost unit
function of the number of membranes. With a low number of
Biogas cost 16e0.2712 $/MBtu - V/Nm3 membranes, the rise in the LCOH is mainly due to the addi-
Upgrading cost 3 $/MBtu tional feedstock cost. In this region, the HRF is low, and then a
Biomethane cost 19e0.532 $/MBtu - V/Nm3
higher amount of BG/BM is needed to reach the target of
Catalyst 540 kV/m3
Filler 50 kV/m3
100 kg/day. Moreover, more heat is available in the retentate
Electric energy 0.12 V/kWh and then a higher heat exchanger area results from the
Maintenance 2 % of TPC calculation, with an increase in plant cost. By increasing the
Insurance 2.5 % of TPC number of membranes, the effect of the increase of HRF for an
Annual labour cost 30 kV additional membrane is always smaller. The LCOH decreases
CCF 0.16 e
since feedstock cost is reduced, while increases due to the
%TIC 0.65 e
additional cost of membranes. After the minimum, cost of the
%IC 0.14 e
%C&OC 0.15 e membranes progressively overtake the savings in feedstock
CEPCI2020 596.2 e cost.

Comparison between biogas and biomethane

LCOH trend
Trends of LCOH for different pressures, reactor diameters and
number of membranes are reported on the left side of Fig. 8. At
the same reactor diameter, the minimum in LCOH values is
similar at all pressures, but in general higher pressures allows
to reach the minimum at a lower number of membranes,
which is preferred since membranes are a breakable compo-
nent in the reactor. At higher pressure, HRF is higher so less
biogas is required at fixed hydrogen production. On the other
side, higher pressures have lower system efficiency, that
means a higher cost for electricity. The preference in working
at higher pressures allows also to adopt smaller diameters
(gas density reduces) and an increase of the LCOH is associ-
ated with the increased reactor diameter since additional
material will be necessary for its realization and more catalyst
fits into the reactor. However, this advantage is limited since
Comparison in terms of LCOH between BG and BM is shown
on the right side of Fig. 8. For both feeds, the shape of the curve
is the same previously discussed: it has a parabolic behaviour
as a function of number of membranes, and LCOH is higher for
higher reactor diameters. It turns out clearly that BM leads in
any case to a higher LCOH. The minimum is obtained, in BM
case, for a lower number of membranes and a smaller reactor
diameter.
Results for the best design case of both feeds are reported
in Table 7, together with the power balances and the flow rates
of some relevant streams of the system. The composition of
LCOH for the best design identified in BG and BM case is re-
ported in Fig. 9. As expected, using BM as feedstock, the TPC is
reduced due to lower membrane reactor cost and lower cost
for gas compressor. There is also a reduction in electricity
cost, due to the lower electric consumption of the gas

Fig. 8 e On the left, typical trend of LCOH, reported for biogas case, as a function of the number of membranes for different
pressures and reactor diameters. On the right, comparison, at 14 bar, of LCOH curves for biogas and biomethane.
19592 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5

compressor, which is about halved. Nevertheless, the addi-

Table 7 e Comparison of results in the best-design case
tional cost of the feedstock is greater than these reduction
for the different feedstocks.
effect, making in the end BG the preferred choice.
Parameter Biogas Biomethane

Reactor temperature ( C) 550 550 Sensitivity analyses
Retentate pressure (bar) 14 14
Vacuum side pressure 0.1 0.1
The limited TRL of the membrane reactor technology entails
Reactor diameter (cm) 39 37
many uncertainties on the membrane reactor cost; for this
Number of membranes () 99 85
Gas feed (Nm3/h) 26.8 16.2 reason, a set of sensitivity analyses has been performed to
Water flow rate (kmol/h) 1.07 1.11 investigate the impact of membranes, reactor and catalyst
Air (to reactor) flow rate (kmol/h) 1.48 1.56 costs on the LCOH. Reactor cost (without membranes nor
Air (to burner) flow rate (kmol/h) 0.90 0.78 catalyst) is now set to 3.5 times the raw material cost, but it is
BG/BM compressor power (kW) 4.42 2.69 here investigated from 2 to 15 times of material cost. Mem-
Air compressor power (kW) 5.21 5.47
brane cost is investigated if halved or doubled. Catalyst costs
Vacuum pump power (kW) 8.57 8.57
Water pump power (kW) 0.01 0.01
is halved or multiplied by a factor 4. Other parameters
H2 compressor power (kW) 8.36 8.36 investigated are the annual labour cost to 60 kV (currently 30
Thermal power for steam production (kW) 24.38 21.51 kV), also difficult to be evaluated in this stage, and the feed-
Thermal power exchanged HX-4 (kW) 3.21 2.14 stock cost. The latter parameter has been varied in the range
Thermal power exchanged HX-5 (kW) 1.23 1.05 of the BG production cost available [8], since it has been cho-
Total electric requirement (kW) 29.23 27.60
sen for this work an average cost, but it can be very different
Gas power input (kW) 155.0 155.1
depending on the situation. For BG it goes from 0.136 V/Nm3 to
H2 power output (kW) 137.8 137.8
Hydrogen production (kg/day) 100 100 0.364 V/Nm3 (currently 0.2712 V/Nm3) and for BM from 0.266
HRF (%) 97.0 97.4 V/Nm3 to 0.72 V/Nm3 (currently 0.532 V/Nm3). The BM cost is
System efficiency (%) 62.7 63.7 directly related to BG range, obtained by adding to the BG cost
SEC (kWh/kg) 53.2 52.4 a fixed upgrading cost.
LCOH (V/kg) 4.39 4.62 Dealing with vacuum hydrogen requires high-grade steels
LCOH (V/MWh) 132.7 139.5
components for which still hold uncertainties in their costs. A

Fig. 9 e On the left, LCOH comparison for BG vs BM divided in its components; on the right, TPC composition for BM (top) and
BG (bottom).
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
Fig. 10 e Sensitivity analysis on LCOH for some relevant parameters.
sensitivity analysis has been then performed to investigate
the influence of components cost on vacuum section (vacuum
pump, heat exchangers ECO and SH and the hydrogen
compressor) on final LCOH calculated in the two solutions.
The cost is investigated up to 5 times the value calculated by
correlations, and the relative advantage of biogas over bio-
methane is confirmed, with a LCOH increase up to 5.03 V/kgH2
and 5.31 V/kgH2 respectively. Similarly, an analysis on TPC
influence over total cost has been performed by due to further
possible under-estimations. Doubling TPC will end up in a
final LCOH of 5.12 V/kg for biogas and 5.30 V/kg for bio-
methane. Again, the relative advantage of biogas over bio-
methane holds. In this case, the share of TPC on final LCOH
goes from 13% to 22%.
The sensitivity analysis, which results are presented in
Fig. 10, shows again the important impact of the feedstock
cost in the final LCOH. Effect of costs of the MR components is
instead limited. Regarding the shape of LCOH charts and the
new optimum design point, changes in feedstock cost or in
labour cost or in catalyst cost has no influence, so the opti-
mum point is always at 14 bar and for the diameter and
number of membranes listed in Table 7, depending on BG or
BM cases. Also in the case of an increase of the reactor cost,
the optimum remain the same in terms of pressure and
number of membranes. Regarding the diameter, it is relevant
to state that in this case the influence is higher than in the
reference case considered: with 15 times the raw material
cost, generally an increase of 1 cm in reactor diameter leads to
an increase of 0.01 V/kg in LCOH. For the membranes, a vari-
ation in the cost has also an influence on the shape of LCOH
curves: considering a double cost, the minimum moves to-
wards lower number of membranes (90 for BG, 75 for BM); if
the cost is halved, minimum LCOH is obtained at 105 mem-
branes in BG case while it stays constant at 85 membranes in
BM case. Optimum pressure is, in all cases, 14 bar.
Maintaining the same parameters used in this analysis, it is
also possible to estimate what should be the average cost of
upgrading to reach the same LCOH. It turned out that it should
be 2.84 V/MWh (0.97 $/MBtu), so a reduction of about 70%
compared to the current average cost, which is 8.79 V/MWh (3
$/MBtu).
19593
Conclusions
This work presented a tecno-economic assessment of a
small-scale hydrogen production plant using membrane
reactor technology. The analyses have been performed for
different pressures, reactor diameters and number of mem-
branes with the aim of investigating the trends in the effi-
ciency parameters, both from the point of view of the reactor
(HRF) and of the overall system (hsystem ). Moreover, a com-
parison between biogas and biomethane as feedstocks has
been performed. Through an economic evaluation has been
identified the best design of both cases, as the one which
minimize the LCOH.
From a thermodynamic point of view, it turned out that
both HRF and system efficiency increase by increasing the
number of membranes: the growth is sharper for lower
membranes and reaches a plateau around 115e125 mem-
branes. Higher pressures outperform lower pressure in terms
of reactor efficiency but are less efficient from the point of
view of the overall system. The asymptotic values are higher if
biomethane is used as feedstock, due to lower inert (CO2)
presence which increases hydrogen partial pressure and thus
driving force for permeation. Moreover, with lower molar flow
within the reactor, the fluidization velocity decreases and this
positively affects efficiency of the reactor since residence time
increases.
From an economic point of view, given a certain feed-
stock, a higher pressure leads to minimum in LCOH at lower
number of membranes. Using BM, the optimum configura-
tion in terms of LCOH allows to work with smaller reactor
and a less membranes compared to biogas, thus to reduce
TPC and fixed O&M. Nevertheless, even if the amount of
feedstock is reduced, with the current upgrading cost, biogas
results the most convenient choice in terms of LCOH, since
the higher feedstock cost exceeds savings in TPC and auxil-
iaries consumptions. It is anyhow important to stress that,
since MRs are still being validated at TRL7, economic as-
sumptions contain several uncertainties, that will be possibly
refined along with the industrial development of the
technology.
19594 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5

[14] Brencio C, Fontein FWA, Medrano JA, Di Felice L, Arratibel A,

Declaration of competing interest Gallucci F. Pd-based membranes performance under
hydrocarbon exposure for propane dehydrogenation
The authors declare that they have no known competing processes: experimental and modeling. Int J Hydrogen
financial interests or personal relationships that could have Energy 2022;47:11369e84. https://doi.org/10.1016/
j.ijhydene.2021.09.252.
appeared to influence the work reported in this paper.
[15] Marra L, Wolbers PF, Gallucci F, Annaland MVS.
Development of a RhZrO2 catalyst for low temperature
autothermal reforming of methane in membrane reactors.
Acknowledgments Catal Today 2014;236:23e33. https://doi.org/10.1016/
j.cattod.2013.10.069.
[16] Wang S, Luo K, Hu C, Sun L, Fan J. Impact of operating
MACBETH has received funding from the European parameters on biomass gasification in a fluidized bed
reactor: an Eulerian-Lagrangian approach. Powder Technol
Union's Horizon 2020 research and innovation program (grant
2018;333:304e16. https://doi.org/10.1016/
agreement No 869896).
j.powtec.2018.04.027.
[17] Medrano JA, Tasdemir M, Gallucci F, van Sint Annaland M.
On the internal solids circulation rates in freely-bubbling
references gas-solid fluidized beds. Chem Eng Sci 2017;172:395e406.
https://doi.org/10.1016/j.ces.2017.06.046.
[18] Helmi A, Voncken RJW, Raijmakers AJ, Roghair I, Gallucci F,
van Sint Annaland M. On concentration polarization in
[1] IEA. Net zero by 2050: a roadmap for the global energy sector.
fluidized bed membrane reactors. Chem Eng J
Int. Energy Agency 2021;224. Net Zero by 2050 e Analysis -
2018;332:464e78. https://doi.org/10.1016/J.CEJ.2017.09.045.
IEA
[19] Gallucci F, Van Sint AM, Ae A, Kuipers JAM, Van AM, 
[2] Electricity. World energy outlook 2019 e analysis - IEA (n.d.),
Annaland S. Autothermal Reforming of Methane with
https://www.iea.org/reports/world-energy-outlook-2019/
Integrated CO 2 Capture in a Novel Fluidized Bed Membrane
electricity.
Reactor. Part 2 Comparison of Reactor Configurations
[3] International Energy Agency. The Future of Hydrogen:
Number of CSTRs in the bubble phase N e Number of CSTRs
seizing today's opportunities. IEA Publ.; 2019. p. 203. https://
in the emulsion phase. Top Catal 2008;51:146e57. https://
www.iea.org/reports/the-future-of-hydrogen.
doi.org/10.1007/s11244-008-9127-7.
[4] Global hydrogen review 2021. Glob. Hydrog. Rev. 2021;2021.
[20] Di Marcoberardino G, Gallucci F, Manzolini G, van Sint
https://doi.org/10.1787/39351842-en.
Annaland M. Definition of validated membrane reactor
[5] Nikolaidis P, Poullikkas A. A comparative overview of
model for 5 kW power output CHP system for different
hydrogen production processes. Renew Sustain Energy Rev
natural gas compositions. Int J Hydrogen Energy
2017;67:597e611. https://doi.org/10.1016/j.rser.2016.09.044.
2016;41:19141e53. https://doi.org/10.1016/
[6] Clark D, Malerød-Fjeld H, Budd M, Yuste-Tirados I, Beeaff D,
J.IJHYDENE.2016.07.102.
Aamodt S, Nguyen K, Ansaloni L, Peters T, Vestre PK,
~ ana S, Norby T, [21] Foresti S, Di Marcoberardino G, Manzolini G, De Nooijer N,
Pappas DK, Valls MI, Remiro-Buenaman
Gallucci F, van Sint Annaland M. A comprehensive model of
Bjørheim TS, Serra JM, Kjølseth C. Single-step hydrogen
a fluidized bed membrane reactor for small-scale hydrogen
production from NH3, CH4, and biogas in stacked proton
production. Chem. Eng. Process. - Process Intensif.
ceramic reactors. Science 2022;376:390e3. https://doi.org/
2018;127:136e44. https://doi.org/10.1016/j.cep.2018.01.018.
10.1126/science.abj3951.
~ iga J, [22] Di Marcoberardino G, Vitali D, Spinelli F, Binotti M,
[7] Fernandez E, Coenen K, Helmi A, Melendez J, Zun
Manzolini G. Green hydrogen production from raw biogas: a
Pacheco Tanaka DA, Van Sint Annaland M, Gallucci F.
techno-economic investigation of conventional processes
Preparation and characterization of thin-film Pd-Ag
using pressure swing adsorption unit. Processes 2018;vol. 6.
supported membranes for high-temperature applications.
https://doi.org/10.3390/pr6030019.
Int J Hydrogen Energy 2015;40:13463e78. https://doi.org/
[23] Di Marcoberardino G, Foresti S, Binotti M, Manzolini G.
10.1016/j.ijhydene.2015.08.050.
Potentiality of a biogas membrane reformer for decentralized
[8] International Energy Agency, Outlook for biogas and
hydrogen production. Chem. Eng. Process. - Process Intensif.
biomethane. Prospects for organic growth. World Energy
2018;129:131e41. https://doi.org/10.1016/j.cep.2018.04.023.
Outlook Special Report.; 2020. p. 93. https://www.iea.org/
[24] Spallina V, Pandolfo D, Battistella A, Romano MC, Van Sint
reports/outlook-for-biogas-and-biomethane-prospects-for-
Annaland M, Gallucci F. Techno-economic assessment of
organic-growth.
membrane assisted fluidized bed reactors for pure H2
[9] Kapoor R, Ghosh P, Kumar M, Vijay VK. Evaluation of biogas
production with CO2 capture. Energy Convers Manag
upgrading technologies and future perspectives: a review.
2016;120:257e73. https://doi.org/10.1016/
Environmental Science and Pollution Research; 2019. https://
J.ENCONMAN.2016.04.073.
doi.org/10.1007/s11356-019-04767-1.
[25] Brencio C, Maruzzi M, Manzolini G, Gallucci F. Butadiene
[10] Scarlat N, Dallemand JF, Fahl F. Biogas: developments and
production in membrane reactors: a techno-economic
perspectives in europe. Renew Energy 2018;129:457e72.
analysis. Int J Hydrogen Energy 2022;47:21375e90. https://
https://doi.org/10.1016/j.renene.2018.03.006.
doi.org/10.1016/j.ijhydene.2022.04.259.
[11] Di Marcoberardino G, Binotti M, Manzolini G, Viviente JL,
[26] Etip bioenergy. Biomethane fact sheet (n.d.), https://www.
Arratibel A, Roses L, Gallucci F. Achievements of European
etipbioenergy.eu/fact-sheets. [Accessed 7 June 2022].
projects on membrane reactor for hydrogen production. J
[27] de Nooijer N, Sanchez JD, Melendez J, Fernandez E, Pacheco
Clean Prod 2017;161:1442e50. https://doi.org/10.1016/
Tanaka DA, van Sint Annaland M, Gallucci F. Influence of
j.jclepro.2017.05.122.
H2S on the hydrogen flux of thin-film PdAgAu membranes.
[12] BIONICO project (n.d.), http://www.bionicoproject.eu/.
Int J Hydrogen Energy 2020;45:7303e12. https://doi.org/
[13] MACBETH project (n.d.), https://www.macbeth-project.eu/.
10.1016/j.ijhydene.2019.06.194.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 8 ( 2 0 2 3 ) 1 9 5 8 0 e1 9 5 9 5
[28] Kato K, Wen CY. Bubble assemblage model for fluidized bed
catalytic reactors. Chem Eng Sci 1969;24:1351e69. https://
doi.org/10.1016/0009-2509(69)85055-4.
[29] Kunii D, Levenspiel O. Fluidization engineering. 2nd ed. 1991.
https://doi.org/10.1016/0032-5910(93)87011-c.
[30] Numaguchi T, Kikuchi K. Intrinsic kinetics and design
simulation in a complex reaction network; steam-methane
reforming. Chem Eng Sci 1988;43:2295e301. https://doi.org/
10.1016/0009-2509(88)87118-5.
[31] Voncken RJW, Roghair I, van Sint Annaland M. A numerical
study on concentration polarization in 3D cylindrical fluidized
19595
beds with vertically immersed membranes. Chem Eng Sci
2019;205:299e318. https://doi.org/10.1016/j.ces.2019.05.010.
[32] Gas for Climate. Market state and trends in renewable and
low-carbon gases in Europe. 2020 14. https://www.
europeanbiogas.eu/wp-content/uploads/2020/12/GfC_
MSTReport_2020_final.pdf.
[33] 2020 annual CEPCI average value - chemical Engineering |
Page 1, (n.d.), https://www.chemengonline.com/2020-
annual-cepci-average-value/.